Ole Petter Ronningen
2018-05-14 20:15:34 UTC
Perhaps this is of interest to some. I finally got around to doing an
experiment I've been mulling over for some time.
In short, I use a Novatel OEMV-3 GPS receiver, clocked by an Oscilloquartz
8600 BVA, process the observations with RTKLib and real time corrections
from igs - and use the resulting clock solutions as inputs to a software
PLL that sets the EFC on the BVA via a Keysight 33510B function generator
set to DC. A bit of a contraption, but it works surprisingly well (or
perhaps I should say it surprisingly works!)
I measured the GPSDO using a TimePod and my maser.
There is *much* room for improvement, and probably too little data, but I
thought the preliminary results worthy of a post: The ADEV is <5e-13 on the
"hump". It passes through to e-14 territory around 6-7000 seconds, at least
in this dataset. The TDEV flattens out around 2e-10 at 2000 seconds or so
(a little wiggle on the end, but I choose to ignore that and blame it on
not enough data).
If the masks in TimeLab are to be trusted, it is very close to a 5071A high
performance option in this region (400 to 10000 seconds), and of course
much better at lower taus.
The frequency accuracy stays within 2-3e-12, and the phase rarely wanders
off more than 1 ns.
I have pretty much everything to learn about tuning GPSDO's, but I think it
shows that the approach is viable. I ofcourse also need to collect a *lot*
more data, get a proper DAC in place and so forth, but still.
To get some sort of baseline, I also processed in real time using the same
correction stream observations from a Trimble NetRS clocked by the hydrogen
maser (sadly not at the same time) - the resulting curves should pretty
accurately show the limits of the approach - at least using the corrections
I was using (CLK93; there are many to chose from, and frankly I don't know
which would be best for something like this. )
The .tim-files are here: http://www.efos3.com/downloads/GPSDO_14.05.18.zip
As a final thought - this setup can be made "self documenting": log the
observations to RINEX-files, and solve with PPP and IGS Rapid or other
corrections. The resulting phase record should be good to at least low e-15
from what I understand (and observe).